CN-121980164-A - Bank outlet consumable abnormal use root cause diagnosis method, system and computer readable storage medium

CN121980164ACN 121980164 ACN121980164 ACN 121980164ACN-121980164-A

Abstract

The invention relates to the technical field of data processing, in particular to a method, a system and a computer readable storage medium for diagnosing abnormal use root cause of consumable materials in banking outlets. The method comprises the steps of acquiring data through automatic capturing and collecting system data and manual recording, processing the data in parallel based on business coupling degree analysis, time sequence mode detection and peer efficiency benchmarking and inventory strategy compliance verification, respectively generating business interpretation contribution rate, abnormal mode labels, efficiency grades and inventory compliance labels, carrying out standardized mapping and causal probability reasoning on structural evidences output by the agents through a Bayesian network, calculating posterior probability of each cause hypothesis, screening competitive hypotheses, and generating a structural diagnosis report containing most probable causes, competitive hypotheses, key evidence chains and action guidance according to the posterior probability. The invention can realize multi-dimensional root cause diagnosis of abnormal use of the consumable materials in the banking website, and improves management precision and interpretability.

Inventors

QIAO GUANFENG
WU JING
YIN HONG
DENG LU
WANG XIAOLONG

Assignees

中国建设银行股份有限公司河北省分行

Dates

Publication Date: 20260505
Application Date: 20251222

Claims (10)

1. The method for diagnosing the abnormal use root cause of the consumable at the banking website is characterized by comprising the following steps of: S1, acquiring data by automatically capturing data of a centralized mining system and manually inputting the data, wherein the data comprises consumable use data, early warning rules and inventory parameters; S2, processing the data in parallel based on service coupling degree analysis, time sequence mode detection, peer performance benchmarking and inventory policy compliance verification, and respectively generating service interpretation contribution rate, abnormal mode labels, performance level and inventory compliance labels; s3, carrying out standardized mapping and causal probability reasoning on the structured evidence output by the agent through a Bayesian network, calculating posterior probability of each root cause hypothesis and screening competitive hypotheses; S4, generating a structured diagnosis report comprising the most probable root cause, competitive hypothesis, key evidence chain and action guidance according to the posterior probability.
2. The method of claim 1, wherein the method for processing data based on service coupling degree analysis in step S2 is as follows: A21, constructing a linear model based on the acquired data to construct direct correlation between target website traffic and consumable consumption, and predicting consumable consumption under the direct correlation based on the linear model; A22, constructing a nonlinear model based on the acquired data to construct indirect correlation between target website traffic and consumable consumption, and predicting consumable consumption under the indirect correlation based on the nonlinear model; A23, fusing the consumption of the consumable under direct association and the consumption of the consumable under indirect association according to set weights to obtain theoretical consumption driven by the business volume, and judging residual error abnormality and generating corresponding residual error abnormality scores when the difference between the actual consumption and the theoretical consumption exceeds a preset threshold; A24, based on theoretical consumption, using SHAP attribution analysis to quantify the total contribution of various services to consumption fluctuation, and generating a service interpretation contribution rate.
3. The method of claim 1, wherein the method for processing data based on the timing pattern detection in step S2 is as follows: B21, splitting consumable usage data of a target website into a trend term, a season term and a residual term by STL decomposition, and detecting point abnormality and context abnormality in the residual term by an S-H-ESD algorithm, wherein the point abnormality is the situation that consumable usage number of a single time point suddenly increases, and the context abnormality is the situation that the residual term of continuous days exceeds a threshold value or bursts and is in a business stationary period; and B22, respectively calculating the abnormal intensity of the point abnormality and the context abnormality by combining the duration time of the abnormality and the deviation degree, and outputting a time stamp, an abnormality type and the abnormal intensity corresponding to the abnormality as an abnormality mode label.
4. The method according to claim 1, wherein step S2 is performed by the peer performance-based peer-to-peer processing of data by: C21, based on the acquired data, utilizing a K-Prototypes algorithm to cluster the peers of each website, wherein the cluster characteristics comprise service structure duty ratio, scale level, equipment model and region; 22, measuring the deviation degree of the target network point and the peer group through the MAD index, and measuring the efficacy grade of the target network point according to the deviation degree, wherein the efficacy grade comprises leading, average and lagging; And C23, when the performance grade of the target network node is lagging, analyzing the effect lagging reason of the network node, and outputting the clustering basis, the performance grade of each network node and the lagging reason of the peer group list.
5. The method of claim 1, wherein the method of processing data based on inventory policy compliance verification in step S2 is: D21, calculating theoretical safety inventory and maximum inventory based on a traditional (S, S) inventory model, and introducing a validity period constraint correction formula: Wherein, the In order to be in a safe inventory, Is the maximum stock; d22, generating an inventory compliance tag based on the current inventory, the safety inventory, and the maximum inventory, and after the current inventory is greater than the revised maximum inventory When the time from the last purchase is less than 50% of the early period of purchase and the current stock is greater than the safety stock When this is the case, it is determined that "early purchase" is performed.
6. The method of claim 5, wherein the safety stock is calculated by the formula: Daily consumption x procurement advance period x safety factor; The purchase lead time is determined by the delivery time of the supplier; The calculation formula of the maximum inventory is as follows: Safety stock + 30 days consumption.
7. The method of claim 1, wherein S3 further comprises: S31, converting the generated service interpretation contribution rate, the abnormal mode label, the efficiency level and the inventory compliance label into three-level labels, and mapping the labels into Bayesian network evidence node probabilities; S32, inputting the evidence probability into a network, reasoning through a variable elimination method, and updating the root cause node posterior probability by using a preset causal relation; s33, root nodes with the posterior probability larger than a set probability threshold are screened from the root nodes, the root nodes are ranked according to the posterior probability from large to small, the root node with the highest probability is set as the most probable root sound, and 30% of the root nodes before ranking are used as competitive hypothesis root sounds.
8. The method of claim 7, wherein if the posterior probabilities of all root cause nodes are less than the probability threshold, prompting a worker to supplement a new root cause.
9. A banking site consumable abnormal use root cause diagnosis system, characterized by comprising: the data acquisition module is used for acquiring consumable use data, early warning rules and inventory parameters through automatic grabbing collection system data and a manual entry mode; The intelligent agent processing module is used for processing the data in parallel based on the business coupling degree analysis, the time sequence mode detection, the peer efficacy benchmarking and the inventory policy compliance verification, and respectively generating a business interpretation contribution rate, an abnormal mode label, an efficacy grade and an inventory compliance label; the Bayesian network processing module is used for carrying out standardized mapping and causal probability reasoning on the structured evidence output by the intelligent agent through a Bayesian network, calculating posterior probability of each root cause hypothesis and screening competitive hypotheses; And the diagnosis report generation module is used for generating a structured diagnosis report comprising the most probable root cause, competitive hypothesis, key evidence chain and action guidance according to the posterior probability.
10. A computer readable storage medium storing a computer program which, when executed by a processor, implements the method of any one of claims 1-8.

Description

Bank outlet consumable abnormal use root cause diagnosis method, system and computer readable storage medium Technical Field The invention relates to the technical field of data processing, in particular to a method, a system and a computer readable storage medium for diagnosing abnormal use root cause of consumable materials in banking outlets. Background Along with the comprehensive promotion of the digitalized transformation of banks, although paperless office work is further realized, banking sites are taken as the first line of face customers, a great amount of printing, copying, form filling and other requirements still exist, the related consumable demand is high, meanwhile, in order to ensure the normal business of the sites and improve the customer satisfaction, a relatively extensive management mode is adopted on the consumable management and control, the sites apply for consumable materials at the moment, the purchasing department excessively purchases the consumable materials, and the like, and although the use of the consumable materials is ensured, the use waste condition of the consumable materials is also caused, so that more refined management is needed. However, in the prior art, there is a problem of insufficient fine management of the consumable management method. At present, the declaration, the use and the purchase of consumable materials mainly adopt traditional manual statistics, and the website habit is used as the consumable materials are applied, but the use details, the progress tracking, the excess reminding and the like of various consumable materials are not digitalized and visually managed. Traditional management relies on fixed threshold or single time series models, which only reveal the "out of limits" appearance, and cannot penetrate to the root cause of "why out of limits". Disclosure of Invention The present invention aims to solve at least one of the technical problems in the related art to some extent. Therefore, a first object of the present invention is to provide a method for diagnosing abnormal use of consumable materials in banking sites, comprising: S1, acquiring data by automatically capturing data of a centralized mining system and manually inputting the data, wherein the data comprises consumable use data, early warning rules and inventory parameters; S2, processing the data in parallel based on service coupling degree analysis, time sequence mode detection, peer performance benchmarking and inventory policy compliance verification, and respectively generating service interpretation contribution rate, abnormal mode labels, performance level and inventory compliance labels; s3, carrying out standardized mapping and causal probability reasoning on the structured evidence output by the agent through a Bayesian network, calculating posterior probability of each root cause hypothesis and screening competitive hypotheses; S4, generating a structured diagnosis report comprising the most probable root cause, competitive hypothesis, key evidence chain and action guidance according to the posterior probability. In one embodiment of the present invention, the method for processing data based on service coupling degree analysis in step S2 is as follows: A21, constructing a linear model based on the acquired data to construct direct correlation between target website traffic and consumable consumption, and predicting consumable consumption under the direct correlation based on the linear model; A22, constructing a nonlinear model based on the acquired data to construct indirect correlation between target website traffic and consumable consumption, and predicting consumable consumption under the indirect correlation based on the nonlinear model; A23, fusing the consumption of the consumable under direct association and the consumption of the consumable under indirect association according to set weights to obtain theoretical consumption driven by the business volume, and judging residual error abnormality and generating corresponding residual error abnormality scores when the difference between the actual consumption and the theoretical consumption exceeds a preset threshold; A24, based on theoretical consumption, using SHAP attribution analysis to quantify the total contribution of various services to consumption fluctuation, and generating a service interpretation contribution rate. In one embodiment of the present invention, the method for processing data based on the timing pattern detection in step S2 is as follows: B21, splitting consumable usage data of a target website into a trend term, a season term and a residual term by STL decomposition, and detecting point abnormality and context abnormality in the residual term by an S-H-ESD algorithm, wherein the point abnormality is the situation that consumable usage number of a single time point suddenly increases, and the context abnormality is the situation that the residual term of continuous days exceeds a threshold